Complex problems, disruptive ideas, and predictive models are natural targets for scientific research in all areas of knowledge, whether to become more knowledgeable or to solve problems.
It would be unfair to say that some areas face more challenging problems than others. After all, here we are, researching, inventing, solving, dispelling fog, and expanding horizons. All topics in all areas are challenging.
Maybe some are more wicked than others. Let me explain.
The expression “wicked problems”, disseminated from the 1970s onwards with the work of Rittel and Webber (1973), can be translated into Portuguese as “wicked problems”. They are wicked because they do not have definitive formulations and are not reproducible.
Other adjectives, such as "complex," are also used to describe them. Since almost all research problems are complex, in all fields, I prefer "perverse," because this term is more appropriate for those who want to unravel the workings and propose explanatory models for extremely complex objects.
It's no coincidence that the term "wicked problems" was coined for challenges in applied social sciences, especially political science, economics, and sociology. I'd argue that it applies to management as well.
As Head (2008) explains, wicked problems are “social problems, generally ill-defined, non-categorical, interconnected, and based on judgment rather than scientific evidence.” To make matters more challenging, Rittel and Webber (1973) assert that “every wicked problem is essentially unique.”
As Master Yoda would say, It is complex to find evidence of models that are reproducible.
The purpose of this introduction is to say that the most complex areas of research are precisely those dedicated to this type of problem. I would venture to say that it is in the applied social sciences and the humanities in general that most of the wicked problems arise, which makes scientific research in these areas a huge challenge. As we will see later, if research is framed as interdisciplinary, this condition is enhanced.
At the root of wicked problems is the fact that, to function, models must depend on the behavior and decisions of individuals and groups of individuals living in society. Add to this the uncertainty that naturally arises from the passage of time, and an inevitable asymmetrical state is created between what I say and think today and what I will do the next moment. Whatever the cause of my change of position, it could render a predictive model unfeasible, without an scrutable motivation.
Two correlated elements of uncertainty need to be considered in an explanatory and/or predictive model in these areas: how people will behave and make decisions and the evolution of the variables that will influence those decisions.
To complicate matters, these influencing variables will also be influenced by decisions. Since one depends on the other, the task of modeling is difficult. But we keep trying.
The results will be models that will only eventually be reproducible, precisely if the decisions follow the predicted path, and the interactions between the decisions follow as well. No other important variables can emerge, otherwise the model will fail.
Kay & King's (2020) book, titled Radical Uncertainty: decision making for an unknowable future, helps to understand the chasm that exists between the applied social sciences and the natural sciences. Of course, this isn't to say that the latter don't have complex and perverse problems. They do exist, but for different reasons: in the latter, systems can be closed to demonstrate their functioning. In social systems, this isn't the case. Closing them down is like denying the uncertainty surrounding how individuals and groups think and decide.
The aforementioned book was written by two renowned economists, one of whom served as Governor of the Bank of England for ten years. From the outset, the authors acknowledge that predictive models of macroeconomic variables, even very short-term ones, fail or have low accuracy, despite the increasing incorporation of variables that influence decisions and outcomes. In this sense, they acknowledge that radical uncertainty is insurmountable and that models have limited explanatory and, especially, predictive reach. We have to get used to this, they say.
In one of the book's many interesting passages, the authors recount the decision-making process President Barack Obama reportedly went through in the Oval Office regarding Osama bin Laden's exact location. Was he where his aides believed he was? I reproduce here an excerpt from the book that I find illustrative:
"John, the CIA chief of staff, was 95 percent certain that bin Laden was in the building. But others were less certain. Most estimated the probability at about 80 percent. Some were as low as 40 percent or even 30 percent.
'But that's like flipping a coin,' the president said! 'I can't base a decision on that kind of estimate.'
Obama understood that he needed to make a decision based on limited information…
And he did so not through probabilistic reasoning, but by asking: “What’s really at stake here?” (Kay & King, 2020).
The rest is history.
Limited information. We will always deal with it. The only way to know it all is within a closed system, where everyone's identity and behavior are known. A single "hole" in this system can alter the outcome.
One might argue that, to solve the situation, it would be sufficient to have a flexible model with, for example, a relatively generous sensitivity, above or below expectations. Others might argue that a statistical approach, say, "renewable" at every moment, could approximate the correct result, such as the Bayesian approach.
All of this is done with the purpose of dealing with uncertainty. If this weren't the case, there wouldn't be a need to resort to probabilities with broad sensitivities, as we often do.
In the natural sciences, similar problems are also faced due to a lack of knowledge about variables. Not because of radical uncertainty, but because of a lack of models and variables that remain to be known. After all, they exist in nature. Or, as a physicist once told me, "if they exist, it's because they're already there somewhere."
Sabine Hossenfelder's interesting book, entitled Existential Physics: a scientist's guide to life's biggest questions (Hossenfelder, 2023), discusses this point. I dare not enter into the physicists' discussions about the "theory of everything," who argue that "the future is fixed" and that it could not be otherwise.
However, the author, who repeats this statement several times throughout the book, acknowledges the right and duty of structured skepticism. At one point, she transcribes an interview she conducted with David Deutsch, a physicist involved in quantum computing.
The main question of the interview was precisely the predictability of physical systems and, by extension, of everything else. The interviewee then makes a distinction between being deterministic and being predictive.
The author asks the physicist, "Would you say that everything can be deterministic, not just computers, but also consciousness and human behavior?" He replies, "Yes, everything can be deterministic. The state of things at one moment in time is determined by the state at another moment, adding dynamical laws." But note, he says, "that what occurs in time can only be explained by what occurred before, never vice versa."
Arthur Eddington's arrow of time only goes one way, and that's forward, never backward. At least that's how it is in the living world we live in.
And being deterministic, according to the interviewee, does not mean being predictive, for three reasons.
'The first is that, in quantum mechanics, it is not possible to measure the current state with perfect accuracy; the second is that there is chaos, in the sense that any change of a bit nullifies the predictability of a system whose original state can no longer be fully known; the third is that it is not logically possible to predict the future of knowledge, if we did, it is because it would be possible to obtain it before the moment of prediction.'
Hossenfelder concludes, as a corollary to his colleague's statement: 'If we could predict the advance of knowledge, knowledge would not advance.'
Let us accept this logical knot for now and return to the point I want to highlight, that of the perversity of problems in applied social sciences.
The only way to create an effectively predictive model in these areas of knowledge would be to agree that everyone would make the previously agreed-upon decisions, at the right time, and without adding or subtracting anything substantive from what was agreed upon. More challenging, nothing could emerge from outside. A closed system.
Applied social sciences have long attempted to emulate models from the natural sciences to study social phenomena. Isolating variables and assuming everything else constant is the most common approach. After all, we need systems and models that are as deterministic as possible, even if they are probabilistic.
If everyone acts as agreed, the outcome will be predictable. But even if we could isolate individuals in glass domes, without communication or awareness of what's happening, their minds would still be able to generate something that makes them change their minds.
So, bringing the physicists' discussion here, just because it's deterministic doesn't mean it can be predictive. We'll only know the outcome after the necessary time has passed. Again, it's the arrow of time.
The more we study people, societies, and their behaviors, the more difficult it will be to create deterministic and predictive models. The future will be the combination of decisions made in the present by diverse, often broad and divergent, groups of social actors and economic agents, whose motivations are also influenced by others. It is highly challenging to try to organize this into explanatory models, even if we consider them to be deterministic.
The intrinsic uncertainty of all research is magnified here by the impossibility of controlling subjects' actions. This condition leads applied social sciences to pursue diverse approaches, from those borrowed from the natural sciences (physics and biology are the most commonly used) to those based on case studies.
This characteristic differentiates these areas of knowledge in research financing and communication procedures.
Because they offer a wide range of possibilities for studying the same phenomenon, and because they are subject to interpretative scrutiny by peers, which logically conform to the wide range of methodological preferences and convictions about the functioning of human and social things, it is common for these areas to experience greater rejection of funding proposals by development agencies.
It's also common and well-known that scientific communication in these fields differs from that in the natural sciences. Books, book chapters, and symposium proceedings are much more common than in other fields.
These elements are accentuated – and greatly so – when the area is interdisciplinary.
Let's look at some data obtained from Fapesp.
Between 1998 and 2022, applications for regular grants had average approval rates in applied social sciences 27% lower than the average for natural sciences and engineering. In the interdisciplinary field, this rate is 33% lower. These are the two areas with the lowest approval rates. An explanatory hypothesis is precisely the argument used in this text: broad interpretative and methodological approaches, with relatively low convergence that negates each other.
Here, if proven, the bias – and conviction – of the evaluator (are they really peers?) may have more influence than in other areas.
I even doubt that in these areas it can be said that there is a “normal science”, as proposed by Thomas Khun.
To complete the analysis, let us also look at data on publications resulting from FAPESP funding. In Bin et al. (2022)[1], which analyzed different publication patterns among areas of knowledge for former master's and doctoral scholarship holders, the areas of applied social sciences, humanities and language, language and arts, communicate their findings mainly in books and book chapters in much larger proportions than in other areas.
Perhaps, due to the risk of bias that permeates these disciplines, researchers prefer the path of books and chapters rather than that of journals.
However, as publications in journals are increasingly valued, even among applied social researchers there is an increase in the number of journals to welcome those aligned with certain currents and, at the same time, reject followers of other ideas and belonging to other areas of thought.
It's like the anecdote about the shipwrecked Jew who built two synagogues on a desert island. A visitor arrived on the same island later and asked: why two? The answer came immediately: I pray in this one, I don't set foot in that one.
These are side effects of a model of science that apparently fails to recognize the differences and challenges naturally posed by wicked problems.
Is it or isn't it more difficult to do science in these areas?
This text does not necessarily reflect the opinion of Unicamp.
